In [ ]:

    

import numpy as np
def parseMatStr(s, n):
    mat = np.empty((n, n))
    for i, line in enumerate(s.strip().split("\n")):
        mat[i,:] = np.asarray(list(map(int, line.split())))
    return mat


    

In [ ]:

    

matStr = """
  7  53 183 439 863
497 383 563  79 973
287  63 343 169 583
627 343 773 959 943
767 473 103 699 303
"""
mat = parseMatStr(matStr, 5)


    

In [ ]:

    

def _computeScoreMatrix(m,):
    scores = np.empty_like(m)
    norm = m.shape[0]# * m.shape[1]
    for x in range(m.shape[0]):
        for y in range(m.shape[1]):
            if np.isinf(m[x, y]):
                scores[x, y] = -np.inf
                continue
            # Compute total sum of 'blocked' values
            scores[x, y] = (-(np.sum(m[:,y]) + np.sum(m[x,:]) - 2*m[x, y]) / 1) + m[x, y]
            #print(mat2[x,y], mat[x,y])
    return scores

def computeScoreMatrix(mat):
    # Compute, for each cell, the score of that cell (high = good)
    scores = _computeScoreMatrix(mat)
    # Compute the score matrix_computeScoreMatrix
    #return _computeScoreMatrix(scores)
    return scores

def doStuff(mat):
    m = mat.copy()
    # Find maximum score
    m2 = computeScoreMatrix(m).copy()
    for i in range(5):
        idxs = np.unravel_index(np.argmax(m2), m2.shape)
        print(idxs[0], idxs[1], mat[idxs[0], idxs[1]])
        #m2[idxs[0], idxs[1]] = -np.inf
        # Erase section in matrix
        m2[:,idxs[1]] = -np.inf
        m2[idxs[0],:] = -np.inf

doStuff(mat)


    

In [ ]:

    

np.unravel_index(np.argmax(m2), m2.shape)


    

In [ ]:

    

mat[4, 0]


    

In [ ]: